A MACHINE LEARNING BASED APPROACH TO IDENTIFY KEY DRIVERS FOR
IMPROVING CORPORATE’S ESG RATINGS
Dwijendra Dwivedi 1, Saurabh Batra 2 & Yogesh Kumar Pathak3
ABSTRACT
Investors increasingly non-financial factors as part of their risk analysis process and growth
assessments of corporates. Machine learning (ML) models for predicting ESG scores are an
extremely useful tool to help investors make more informed decisions on their portfolios. Such
a tool with wide-encompassing alternative data can be useful to the investors. The use of such
datasets and machine learning models for ESG ratings can continuously improve the accuracy
and reliability of those models. Using machine learning algorithms to identify key drivers of
ESG ratings is an effective way of improving portfolio performance. Although the current state
of ESG ratings is relatively static, data collection and mapping methodologies are evolving. As
more data becomes available, the noise in ESG factors will become less important.
This unique document provides a machine learning algorithm for predicting an ESG rating
based on a company's financial and non-financial attributes. The financial and non-financial
attributes of corporations are extracted from Moody's Orbis and Ratings from S&P. The
objective here is to predict the ESG rating of companies where the ESG rating is not easily
accessible. At the same time, this approach would allow investors to have a suitable framework
for investments based on ESG ratings. With the latest financial and non-financial disclosure by
a corporate an ESG score can be predicted which can be used to identify its riskiness with a
corresponding increase/decrease of ESG score.
Keywords: ESG Investing, ML model, Socially responsible investing, Sustainable
investing, Green finance.
Received: 24/11/2022
Accepted: 27/03/2023
DOI: https://doi.org/10.37497/sdgs.v11i1.242
1
Kracow University of Economics, (Poland). E-mail: dwivedy@gmail.com Orcid id: https://orcid.org/0000-00017662-415X
2
Delhi University, (India), E-mail: saurabhbatra18@gmail.com Orcid id: https://orcid.org/0000-0002-0939-702X
3
Indian Institute of Management Lucknow (India). E-mail: ykpathak@gmail.com
1
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
1.
INTRODUCTION
The corporate social responsibility (CSR) initiatives and their influence on corporate
financial performance have been actively researched (Francesco et al., 2018). The term
encompasses how the company contributes to the preservation of the environment, for instance,
by reducing air pollutant emission, issuing green bonds, and participating in other activities,
which can help solve the problem of climate change. Moreover, environmental, social, and
governance (ESG) initiatives cover social issues, such as caring for employees by providing
good work conditions, effective human capital management, and health and safety programs.
Additionally, corporate governance is implied in ESG activities. ESG refers to the three central
factors in measuring the sustainability and societal impact of a company or a business. These
criteria have proven to be influential on performance due to increased scrutiny on companies to
comply with sustainable business models. The central factors are as below:
Environmental:
Environmental considerations include the contribution of a business or government to
climate change through greenhouse gas emissions, as well as waste management and energy
efficiency. As a result of renewed efforts to address global warming, it is becoming increasingly
important to reduce emissions and decarbonize.
Social
Social aspects include human rights, labor standards in the supply chain, any exposure to
illegal child labor, and more common issues such as respect for workplace health and safety. A
social score is impacted positively if a company is well integrated with its local community
and therefore has a ‘social license’ to operate with consent.
Governance
Governance means a set of rules or principles that define the rights, responsibilities, and
expectations between the different stakeholders in corporate governance. A well-defined
corporate governance system can be used to balance or align interests across stakeholders and
can be used as a tool to support a business's long-term strategy.
ML models have proven to be highly effective in predicting ESG Ratings. They can be
trained on extensive structured data with a range of features. Using a model for ESG will help
2
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
investors and analysts make informed decisions. In addition to predicting ESG ratings, it can
also predict the future performance of a company (Valeria et al., 2021).
One approach to build an ML model for ESG ratings is to collect two weeks’ worth of
news data from various sources. This is computationally expensive because it includes over
37,000 different news articles. Then, the machine learning algorithm produces predictions
based on these features. This model contains positive and negative prediction rules. Each rule
contains two features and a range of time intervals. The second step is to determine whether the
feature should be analyzed for all stocks or just a particular one.
A key problem in analyzing ESG data is the lack of a standard taxonomy for the different
ESG factors. This means that data providers must make model selections and must control for
factors such as country, sector, and risk. These factors affect the prediction accuracy of ESG
ratings and pillar scores. However, machine learning algorithms can help to reduce noise and
improve the accuracy of ESG ratings and pillar scores. This can help sustainable investors to
identify companies that may not have the data needed to meet sustainability commitments.
To develop an ML-model for ESG ratings, we have extracted different data components
that make up a company’s ESG profile. Then, we have attempted to create various ML-based
predictive models based on these attributes and combine it with a corresponding index. As a
next step we compared various models to existing ESG scores and make an informed
investment decision based on those. This also gives us an indication of the champion model
that can be selected from all other models. This model gives us a clear idea of how different
companies perform in an industry/sector.
By the time we have built a champion ML model, we also have a clearer idea of which
attributes are important that will be included in the final model. This robust ESG-ML model
provides a meaningful ESG rating.
The data derived from these sources can be standardized to allow for more accurate and
meaningful predictions. ML models can be useful in predicting ESG ratings; at the same time,
they can be applied to questions related to ESG. In addition to ESG rating, these ML models
can be used for responsible investment in a specific area, such as environmental issues. The
results show that a particular model can accurately identify which stocks are more sustainable
and which are less. The data analyzed in this way will help investors make an informed decision
regarding which ESG stocks to buy. They will also help managers improve their ESG
Reporting.
In predicting ESG ratings, the data should encompass a suitable and comprehensive peer
universe. It should also cover the entire spectrum of ESG. Only then these models should be
3
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
able to provide a consistent and credible representation of a company. However, the social pillar
of ESG has some gaps. Some of the companies have not published all of their ESG reports.
This leads to gaps in the model.
Corporates should have an internal ratings system for ESG and should not rely
compositely on external agency. There is variation in data providers' ESG ratings. The lack of
standardization and lack of transparency in these data sources makes it difficult for companies
to use these metrics to make decisions. Regardless, ESG metrics have a wide range of uses,
from measuring a company's environmental performance to influencing investment decisions.
These data providers use different methodologies to measure ESG, which can result in wide
variations in scores. ESG ratings are complex and have limited correlations. It is difficult to
accurately evaluate the correlation between ratings when they vary greatly, especially when
there is no standardization across the providers. There are three factors that contribute to the
variation: scope, measurement, and weights. These factors are interconnected and can lead to
different results. So a robust internal rating model that is calibrated periodically is the need of
the day and has been tested in the paper.
Non-financial data for ESG rating models can be used to make better rating models and
hence better investment decisions. There are numerous data providers who supply nonfinancial data for ESG rating models. These providers offer a range of data including private
and quasi-public data. They also sell raw data. The data providers' methodological choices are
crucial to the quality of their ratings. Some providers use in-house statistical models to estimate
data that is not reported. One of the biggest challenges of ESG scoring is the sheer size of the
input variables. There is a huge range of variables to consider, from the volume of coverage to
the influence of the source. Here we test various non financial data for building the ratings
models in the paper.
2. LITERATURE STUDY
Valeria et al. (2021) used ESG scores from Bloomberg to study the role of structural
variables adopting a machine learning approach. They used the balance sheet data for a sample
of the Euro Stoxx 600 index constituents and studied how this data explains the Bloomberg
ESG scores. They found that financial statement elements provide a powerful tool for
explaining the ESG rating using machine learning. In another study (Recoo Ciciretti et al.,
2022), they examined whether the ESG investment reduces the risk of contagion between equity
4
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
mutual funds. They measured the impact of fire-sale spillovers, spreading across the financial
system, on funds ranked on ESG aspects. In the event of de-leveraging of funds, the fallout
from the fire sale spreads throughout the network because of the mutual assets of the funds.
They observed that the vulnerability of funds to contagion decreases as the level of ESG
compliance increases.
Furthermore, they found that the average relative loss is lower for funds classified higher
than for funds classified low. Xian Lin et al. [2022] has worked on the application of machine
learning models for predictions on cross-border merger and acquisition decisions with ESG
features from a sustainable development and ecosystem perspective. They used AdaBoost to
train several low-ranking classifiers to achieve a robust decision-making model with a large
financial transaction database of 215,160 transactions. The results achieved an 80.1% prediction
accuracy, using the AdaBoost model, thanks to a 10-step cross-validation. They found
differences between the predictive features of mergers and acquisitions (M&A) and the
different characteristics of sustainable development. Caterina De Lucia et al.(2020) has
developed a machine learning model for publicly owned enterprises in Europe. Their findings
suggest that ML accurately predicts key metrics such as Return on Assets (ROA) and Return
on Equity (ROE), and indicate, through the ordered logistic regression model, the existence of
a positive relationship between ESG practices and the financial indicators. They also found that
the existing relationship becomes more apparent when firms invest in environmental
innovation, employment, productivity, diversity and equal opportunities policies. In a different
work by Carmine de Franco et al. (2018), they designed an ML algorithm which identifies the
models between ESG profiles and financial performance for companies in a broad investment
universe. The algorithm consisted of a set of regularly updated rules that map regions to the
ESG characteristics of high-dimensional space for excess yield prediction. The final predictions
are converted into scores. These scores are used to design simple strategies that filter the
investing universe for equities with a positive rating.
Ng and Rezaee (2020) investigated whether and how business sustainability performance
and disclosure factors affect stock price informativeness (SPI). Fiaschi et al. (2020) conducted
a study to evaluate and overcome the limitations of ESG scores. Rajesh et al. (2021) found in
his research, the mean differences in the CSR strategy scores and the ESG scores of firms in
select developed economies such as US, UK, Japan, and Australia, representing different
geographical regions globally. Champagne et al. (2021) examined whether non-financial ratings
are related to the likelihood of adverse environmental, social, and governance (ESG) events
occurring and therefore serve as an indicator of ESG risk. Bannier et al. (2022) studied the
5
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
relationship between corporate social responsibility and credit risk for U.S. and European firms
over the period 2003 to 2018 and found that only environmental aspects are negatively related
to various measures of credit risk for U.S. firms. For European companies, environmental and
social aspects are adversely associated with credit risk. Jarjir et al. (2020s) ; concluded that a
risk premium is associated with market-assessed non-financial ratings (i.e., environmental,
social and governance (ESG) ratings). You and Managi (2022) investigated whether the
information disclosure of Environmental, Social, Governance (ESG) criteria is more crucial
than actions for the financial performance of firms by using two different ratings with more
than 1,000,000 samples and results showed that disclosure is more important for profits while
action is more critical in Tobin’s Q and Intangible Value Assessment (IVA) scores. Wong et
al. (2021) showed that ESG certification reduces a company's cost of capital, while Tobin's Q
certification increases substantially. Broadstock et al. (2020) the study provided evidence of a
non-linear relationship between the adoption of ESG policies and the innovativeness of
enterprises. Chen and Yang (2020) revealed that investors respond excessively to
environmental rather than social or governance factors. Li et al. (2022) concluded that higher
ESG scores, reduce the risk of business failure. Barros et al. (2022) investigated that whether
mergers and acquisitions (M&A) operations impact firms’ performances on triple ESG pillars
(environment, social, and governance) and provided evidence that M&A deals have a positive
impact on the ESG score of firms. Sabbaghi (2022) provided empirical studies of asymmetric
volatility in ESG holdings. Wong and Zhang (2022) found that Empirical results advance
signalling theory and resource-based view by providing evidence that corporate reputation is
considered a valuable intangible asset by investors and adverse ESG disclosure via media
channels have a significant and negative impact on firm valuation. Avramov et al. (2021)
analyzed asset pricing and the portfolio impact of a significant impediment to sustainable
investment: uncertainty around the company's ESG profile.
Targeted studies were carried out on a smaller group of corporate entities. Egorova et al.
(2022") Demonstrated that enterprises in the Information Technology (IT) sector have
significant weaknesses in their ESG components and reporting. Huang et al. (2022) has
demonstrated that natural disasters have an impact on the environmental, social and governance
(ESG) disclosure policies of the companies located near disaster areas. They found that
companies with a higher percentage of local institutional ownership are more likely to increase
ESG disclosure following nearby disasters. Feng et al. [2022] studied the relationship between
ESG scores and the risk of falling stock prices, finding a statistically and economically
significant negative relationship for Chinese firms. Singhania and Saini (2022) noted that a
6
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
systematic approach using ESG structures could serve as a reference point for the business
sector in India and lead to sustainable development goals (SDGs). Ng and Rezaee (2020),
Becker et al. (2021) Top of page analysed the impact of the Financial Sustainability Disclosure
Regulation (SFDR) on mutual funds and retail investors in the EU. The results showed that the
allocated funds increase their sustainability rating after the policy was implemented.
3. DATA AND METHODOLOGY
ESG Rating Data: NIFTY100 ESG Index is designed to reflect the performance of
companies within NIFTY 100 index, based on Environmental, Social and Governance (ESG)
scores. The weight of each constituent in the index is tilted based on ESG score assigned to the
company, i.e., the constituent weight is derived from its free float market capitalization and
ESG score. To form part of the NIFTY100 ESG Index, stocks should qualify the following
eligibility criteria: Stocks should form part of NIFTY 100 Companies and should have an ESG
score. However, companies engaged in the business of tobacco, alcohol, controversial weapons
and gambling operations has been excluded.
Sample: For the purpose of the study, a sample of 90 companies included in National
Stock Exchange (NSE) - 100 ESG Index. Of the total companies, GlaxoSmithKline Consumer
Healthcare Ltd has been merged with Hindustan Unilever Ltd (HUL) and the data of Cummins
India Ltd, Emami Ltd., Oil India Ltd. Steel Authority of India Ltd (SAIL), and Vodafone Idea
Ltd was not available from the sources of data collection. Therefore, effectively the sample of
the study constitutes 84 companies. The data have been collected for the year 2021.
Sources of Data: ESG Disclosure scores have been collected from www.s&pglobal.com.
The S&P Global ESG Scores is an environmental, social and governance data set that provides
company level, dimension level, and criteria level scores based on the S&P Global Corporate
Sustainability Assessment (CSA) process, an annual evaluation of companies' sustainability
practices. data pertaining to women's participation on board and in Key executive position has
been collected from www.goodreturn.in. Data for financial ratio and other corporate attributes
was extracted from Moody’s Orbis. Moody’s Orbis contains information on companies across
the world and focuses on private company information containing information on companies in
comparable formats. It has information on close to 400 million companies and entities across
the globe. 41 million of these have detailed financial information.
Around 47 attributes were available in Orbis. Key fields extracted from Orbis include:
7
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
• Operating revenue
• Number of employees
• P/L before tax
• Cash flows
• Total Assets
• Current Ratio
• ROE
• Number of publications
• BVD Independence Indicator: This field contains the BVD independence indicator which
indicates independence with respect to shareholding as follows:
1. A+: Indicator A+ identifies Independent Companies and it is attached to any company with
known recorded shareholders, none of which having more than 25% of direct or total
ownership
2. B+: Indicator B+ is attached to any company with a known record shareholder none of
which with an ownership percentage (direct, total or calculated total) over 50% but with
one or more shareholders with ownership percentages above 25%
3. C+: Indicator C+ is attached to any company with a recorded shareholder with a total
or a calculated total ownership over 50%.
4. D: Indicator D is allocated to any company with a recorded shareholder with a direct
ownership of over 50%
5. U: Companies with an unknown degree of ownership concentration.
3.1. Data Preparation:
Data for 84 corporates was extracted from S & P and Moody’s Orbis. Around 47 corporate
level attributes were available. Single Factor analysis was carried out. Pearson Correlation and
VIF score were calculated to check for multicollinearity. Finally, 8 variables were shortlisted
and included in the final Machine Learning model. The final correlation matrix for the selected
variables is as follows:
8
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
Table 1: Correlation Matrix
PL
1
PL
NPUB
NPUB
0.1306
OPREV
0.4993
NCURRDIR
0.1977
NWOMEN
-0.0421
PRMAR
0.1939
NCOMP
0.1582
Score
0.1685
1
0.4381
0.1047
0.0827
-0.2708
0.0044
0.2801
1
0.1451
0.1459
-0.1821
-0.1571
0.2209
1
0.0319
0.1218
-0.1023
0.2799
1
-0.0474
-0.2969
0.3008
1
0.0146
-0.2148
1
-0.1879
OPREV
NCURRDIR
NWOMEN
PRMAR
NCOMP
Where,
PL= P/L before tax USD Last avail. yr,
NPUB=Number of publications,
OPREV= Operating revenue (Turnover) (m USD),
NCURRDIR= Number of current directors and managers,
NWOMEN= Number of women directors,
PRMAR= Profit Margin (%),
NCOMP=Number of companies in corporate group,
Score=ESG Score.
ESG score is positively correlated to all the selected variables except Profit Margin and
number of companies in the corporate group. This behavior is intuitive. The increase in profit
margin will not translate into improvement of the ESG score. Also, as the increase in the number
of subsidiaries in a group of companies, it is difficult to have corresponding increase ESG as
synergies with time.
With respect to the BVD independence indicator, average ESG distribution is as follows:
With the decrease in independence from A+ to D the ESG Score also decreases with the
exception of BVD Grade C+ for which ESG score is less than D. This exception could be
because of small sample size of corporates in grade C+(6) as compared to grade D(27).
9
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
Figure 1: Distribution of BVD independence grade
Machine learning techniques are used to develop the ESG Score Model. Multiple machine
learning model viz. Gradient boosting, neural network, forest, decision tree, linear regression
and ensemble of these models were developed using the shortlisted variables. For the final
model, Gradient Boosting was selected based on the goodness of fit statistic (RMSE).
4. RESULTS
Due to small sample size k-fold cross validation was performed (with k=5). A highly
predictable model is developed with an RMSE of 3.47. The comparison can be summarized in
the tables as follows: The gradient Boosting model gave the best fit with an RMSE of 3.47. The
variable importance of the model is as follows:
Table 2: Model Comparison
Model
RMSE
Gradient Boosting
3.47
Ensemble model
16.12
Neural Network
18.05
Forest
19.15
Decision Tree
20.22
Linear Regression
22.59
10
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
Table 3: Variable importance for gradient boosting model
5.
Variable
Importance
Operating revenue (Turnover) (m USD)
1
Profit Margin (%)
2
P/L before tax USD Last avail. year
3
Number of publications
4
Number of companies in corporate group
5
Number of women directors
6
BVD Independence Indicator
7
Number of current directors and managers
8
CONCLUSION
Several organizations have turned to artificial intelligence (AI) systems to overcome the
challenges involved with analyzing ESG data. Artificial Intelligence can enable companies to
evaluate risks, find hidden opportunities, and perform complex tasks. It can also allow
sustainable investors to filter through mountains of data to identify opportunities and potential
risks.
The author successfully developed a machine learning model to predict the ESG score for
corporate entities. An RMSE value of 3.47 using Gradient boosting algorithm shows high
goodness of fit. This approach of predicting ESG scores for other corporates where such scores
are not available can be used while making ESG investments. Based on the predicted ESG
score, a strategy can be developed using a cutoff value of such score.
ESG ratings are not meant to be used as an indicator of a company's true performance.
However, they can influence a company's decision making and reputation. These ratings can
help companies become more environmentally and socially sustainable, contributing to a more
socially just economy. ESG ratings are becoming increasingly important to investors,
particularly those interested in sustainable economic growth. Since the data for analysis is based
on information from publicly available sources, including annual reports and other disclosures,
the scoring model is easy to use. In addition to helping investors determine whether companies
are performing well on ESG metrics, these reports provide information on competitors and the
11
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
environment. MSCI rates companies on ten-point scales for environmental, social, and
governance factors. This includes controversial business activities such as tobacco production
and weapons manufacturing. The weightings are applied to the data to derive an overall ESG
rating. The underlying methodology behind ESG measures is highly complex, but the results of
this analysis are useful and provide a more nuanced picture of a company’s performance. There
are some drawbacks of ESG ratings, though. One common drawback is the lack of consistency
among rating providers. Some ESG provider score companies differently, and this creates
uncertainty for investors. Additionally, ESG scores are not a complete picture. ML based ratings
models help us to uncover some of these facts.
Machine learning algorithms can be used to analyses data and measure the performance
of companies. They can identify indicators that contribute to efficient portfolios. They can also
measure the performance of companies by comparing their performance with their peers. In
addition, these algorithms can help companies to determine whether their ESG programs are
effective. They can also help to identify companies that may not have the data needed to meet
sustainability commitments.
A common concern among investors is the lack of consistency in the data. For instance,
a company may have a lower ESG score than its peers in one industry but have a higher score
in another. Some providers also issue different ratings on the same company. Internal rating
model can help bridge this gap.
6.
DISCUSSIONS
Developing a proactive ESG rating for corporates is critical for the organizations because
of regulatory and financial needs. It involves a series of steps. They must establish an approach
and strategic plan and companies should develop and manage disclosure. Finally, companies
must implement best practices to improve their ESG ratings. Companies need to work with
different stakeholders to design a focused set of ESG targets and indicators. Companies should
also establish a clear process and establish a system for monitoring ratings. This includes a
system for archiving and auditing. A company must participate in multiple ESG ratings. This
would help to drive strategic decision-making and access lower capital costs. It would also help
to improve shareholder relations. Companies should participate in ratings that align with their
ESG goals and investment philosophy. This will also help to build a relationship with the ratings
analyst team. A rating model like the presented in the paper can help organizations improve
the EST ratings.
12
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
Also, such rating monitoring should be monitored timely update. it should not be a
onetime activity. It would also help to ensure that the company is aware of ESG milestones and
updates. It also includes a comprehensive ESG Risk Rating scorecard. This tool can also be
used to develop a formal ESG risk register. We wish to extent out work with more data sources
and re validate the ratings periodically.
As an extension of this model, Machine learning algorithms can also integrate
unstructured data from public companies to form ESG scores. These metrics are used to
evaluate a company's sustainability and are used by investors to decide which companies to
invest in. In addition, the predicted ESG ratings can be used for companies that are not covered
by rating agencies.
REFERENCES
“Avramov, D., Cheng, S., Lioui, A., and Tarelli, A. (2021). Sustainable investing
with ESG rating uncertainty. Journal of Financial Economics”
“Bannier, C. E., Bofinger, Y., and Rock, B. (2022). Corporate social responsibility
and credit risk. Finance Research Letters, 44(November 2020).”
“Barros, V., Matos, P. V., Sarmento, J. M., and Vieira, P. R. (2022). M&A activity
as a driver for better ESG performance. Technological Forecasting and Social Change,
175(November 2021):121338.”
“Becker, M. G., Martin, F., and Walter, A. (2021). Fostering the Power of ESG
Transparency: The Effect of Sustainability Labels on Mutual Funds and Individual
Investors. SSRN Electronic Journal, (November 2021):1–6.”
“Broadstock, D. C., Matousek, R., Meyer, M., and Tzeremes, N. G. (2020). Does
corporate social responsi- bility impact firms’ innovation capacity? The indirect link
between environmental & social governance implementation and innovation performance.
Journal of Business Research, 119(July 2019):99–110.”
“Champagne, C., Coggins, F., and Sodjahin, A. (2021). Can extra-financial ratings
serve as an indicator of ESG risk? Global Finance Journal, (March):100638.”
“Chen, H. Y. and Yang, S. S. (2020). Do Investors exaggerate corporate ESG
information? Evidence of the ESG momentum effect in the Taiwanese market. Pacific
Basin Finance Journal, 63(October 2019):101407.”
“Egorova, A. A., Grishunin, S. V., and Karminsky, A. M. (2022). The Impact of ESG
factors on the performance of Information Technology Companies. Procedia Computer
Science, 199:339–345.”
“Feng, J., Goodell, J. W., and Shen, D. (2021). ESG rating and stock price crash risk:
Evidence from China. Finance Research Letters, (September):102476.”
“Gangi, F., Mustilli, M. and Varrone, N. (2018). The impact of corporate social
responsibility (CSR) knowledge on corporate financial performance: evidence from the
European banking industry. Journal of Knowledge Management.”
13
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
“Fiaschi, D., Giuliani, E., Nieri, F., and Salvati, N. (2020). How bad is your company?
Measuring corporate wrongdoing beyond the magic of ESG metrics. Business Horizons,
63(3):287–299.”
“Huang, Q., Li, Y., Lin, M., and McBrayer, G. A. (2022). Natural disasters, risk
salience, and corporate ESG disclosure. Journal of Corporate Finance, 72(December
2021):102152.”
“Jarjir, S. L., Nasreddine, A., and Desban, M. (2020). Corporate social responsibility
as a common risk factor. Global Finance Journal, (December 2019):100577.”
“Li, H., Zhang, X., and Zhao, Y. (2022). ESG and Firm’s Default Risk. Finance
Research Letters, (Jan- uary):102713.”
“Ng, A. C. and Rezaee, Z. (2020). Business sustainability factors and stock price
informativeness. Journal of Corporate Finance, 64(November 2019):101688.”
“Rajesh, R., Rajeev, A., and Rajendran, C. (2021). Corporate social performances of
firms in select developed economies: A comparative study. Socio-Economic Planning
Sciences, (November):101194.”
“Sabbaghi, O. (2022). The impact of news on the volatility of ESG firms. Global
Finance Journal, 51(De- cember 2019):100570.”
“Singhania, D. M. and Saini, D. N. (2022). Systems approach to environment, social
and governance (ESG): Case of Reliance industries. Sustainable Operations and
Computers, 3(November 2021):103–117.”
“Wong, J. B. and Zhang, Q. (2022). Stock market reactions to adverse ESG disclosure
via media channels. British Accounting Review, 54(1):101045.”
“Wong, W. C., Batten, J. A., Ahmad, A. H., Mohamed-Arshad, S. B., Nordin, S., and
Adzis, A. A. (2021). Does ESG certification add firm value? Finance Research Letters,
39(November 2019):101593.”
“Yoo, S. and Managi, S. (2022). Disclosure or action: Evaluating ESG behavior
towards financial performance. Finance Research Letters, 44(November 2020):102108.”
“Valeria D’Amato , Rita D’Ecclesia and Susanna Levantesi(2021). Fundamental
ratios as predictors of ESG scores: a machine learning approach.
https://link.springer.com/article/10.1007/s10203-021-00364-5.”
“Roy Cerqueti, Rocco Ciciretti, Ambrogio Dalò and Marco Nicolosi(2022).
Mitigating Contagion Risk by ESG Investing. https://www.mdpi.com/20711050/14/7/3805.”
“Xiangjun Hong, Xian Lin, Laitan Fang, Yuchen Gao and Ruipeng Li(2022).
Application of Machine Learning Models for Predictions on Cross-Border Merger and
Acquisition Decisions with ESG Characteristics from an Ecosystem and Sustainable
Development Perspective. https://www.mdpi.com/2071-1050/14/5/2838.”
“Caterina De Lucia, Pasquale Pazienza and Mark Bartlett. Does Good ESG Lead to
Better Financial Performances by Firms? Machine Learning and Logistic Regression
Models of Public Enterprises in Europe.”
“Vincent Margot, Christophe Geissler, Carmine de Franco and Bruno Monnier. ESG
Investments: Filtering versus Machine Learning Approaches https://arxiv.org/abs/2002.07477.”
14
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT
Dwivedi, D., Batra, S., & Pathak, Y. K. (2023). A machine learning based approach to identify key drivers for
improving corporate’s esg ratings. Journal of Law and Sustainable Development, 11(1), e0242.
https://doi.org/10.37497/sdgs.v11i1.242
Grant/Funding
The authors did not receive any funding for the research, authorship and/or publication of
this article.
Declaration of Conflicting Interests
The authors have not declared any potential conflicts of interest with respect to the
research, authorship and/or publication of this article. SAS software has been used for analysis
and model building.
15
São Paulo| v.11, n. 1| pages: 01-15| e0242 |Jan-Jun| 2023.
SDG
JOURNAL OF LAW AND SUSTAINABLE DEVELOPMENT